Aligning Damages of Heavy-Haul Rail From Images Inspected on Different Dates: An Adaptive Feature Matching Method

Rails are vital components of heavy-haul railway infrastructure, playing a crucial role in ensuring the safety and efficiency of train operations. Analyzing the evolution of rail conditions is essential for effective maintenance planning, particularly when it involves aligning the same damages across different dates. However, due to random mileage deviations in inspection images, it is labor-intensive and highly inefficient to align damages manually and analyze its evolution. To address this issue, we propose a damage alignment method based on adaptive feature matching for inspection images. First, to address the challenge posed by truncated features, we propose a solution using feature size matching to actively identify reference features in a vast collection of inspection images. Next, a key feature selection method, based on multijump differences and Gaussian attenuation coefficients, is designed to retain features with high localization value. Then, a shortest path-based feature matching method is developed to align the same reference features across different dates, achieving the alignment and localization of damages on different dates. Finally, the effectiveness of the proposed method is verified using inspection image data collected from a heavy-haul line on two different dates. Experimental results show that our method achieves accurate alignment of damages in images, with the error between the predicted and actual locations of damages within 1.23 meters.